- Vopak (VPK5 TH) +5.6%
- Vopak Plans Distributions Program of ~€1.7b Through 2030
- Nordex (NDX1 TH) +3.4%
- Nordex Sees 2026 Ebitda Margin 8% to 11%, Est. 8.41%
- AUTO1 (AG1 TH) +1.8%
- AUTO1 FY Revenue Beats Estimates
- Adyen (1N8 TH) +1.5%
- Merlin Properties (MEQA TH) +1.2%
- Wienerberger (WIB TH) +1.2%
- Novo (NOV TH) +1%
- Beiersdorf (BEI TH) -1.3%
- Heidelberg Materials (HEI TH) -2.9%
- Heidelberg Materials Sees RCO of €3.4B to €3.75B for 2026 (1)
- Fresenius SE (FRE TH) -3.1%
- Fresenius Sees 2026 Profit Rise on Hospital, Generics Growth
- Vodafone (VODI TH) -3.2%
- Umicore (NVJP TH) -3.5%
- Umicore Holder Groupe Bruxelles Lambert Offers Shares: Terms
DAX:
- Zalando (ZAL TH) +1.2%
- Beiersdorf (BEI TH) -1.1%
- Beiersdorf Cut to Hold at Deutsche Bank; PT 105 euros
- Fresenius SE (FRE TH) -2.7%
- Fresenius Sees 2026 Profit Rise on Hospital, Generics Growth
- Heidelberg Materials (HEI TH) -3.9%
- Heidelberg Materials Sees RCO of €3.4B to €3.75B for 2026 (1)
MDAX:
- AUTO1 (AG1 TH) +3%
- AUTO1 FY Revenue Beats Estimates
- Nordex (NDX1 TH) +3%
- Nordex Sees 2026 Ebitda Margin 8% to 11%, Est. 8.41%
- Thyssenkrupp (TKA TH) +1%
SDAX:
- Verbio SE (VBK TH) +3.7%
- SFC Energy (F3C TH) +3.4%
- SFC Energy Raised to Buy at Berenberg; PT 17 euros
- Kontron (KTN TH) +1.9%
- PVA TePla (TPE TH) +1.8%
- Tonies SE (TNIE TH) +1.6%
- SMA Solar (S92 TH) -2.3%
Jim Ratcliffe’s Ineos explores asset sales as chemical sector struggles continue
Group searches for ways to raise cash, cut costs and reduce its leverage
Ineos is exploring asset sales and has held talks with creditors over refinancing some of its debt pile, as the chemicals empire built by UK billionaire Sir Jim Ratcliffe battles a prolonged industry downturn.
The London-headquartered group is exploring the sale of assets held within its vinyls business Inovyn, according to people familiar with the matter. The assets could be worth hundreds of millions of pounds, the people said, cautioning that talks remained in preliminary stages.
Ineos has also been in talks with credit firms to refinance bonds that will mature next year, according to several people familiar with the group’s financing situation, while shareholders in recent weeks injected €200mn of new equity into the business alongside raising another €300mn of financing linked to its inventory.
It comes as Ineos searches for ways to raise cash, cut costs and reduce its leverage, as it faces an extended downturn in the chemicals sector.
Europe’s chemicals producers are struggling with a deluge of cut-price imports from China, high energy prices and stifling regulation, on top of overcapacity and weak demand across the globe.
Ineos borrowed heavily as it became one of the world’s biggest chemicals companies through a series of acquisitions. A prolific user of European high-yield debt markets, the group’s bonds have plunged amid wider trouble in the chemicals sector and concern over the company’s leverage.
In 2024, Inovyn was the biggest business by earnings under Ineos Quattro, one of two holding groups the chemicals giant uses to issue debt that is being watched closely by high-yield credit investors.
Debt issued by Quattro had been trading at par in September last year but fell to 65 cents on the dollar last month. It had recovered to above 80 cents on the dollar by mid-February after an equity injection and new financing announced at the end of January.
Quattro’s ratio of net debt to earnings before interest, taxes, depreciation and amortisation (ebitda) stood at 7.7 at the end of 2025, up from 1.9 in 2022. Rating agency S&P regards any company with a net leverage ratio of more than six times to be “highly levered”.
Ratcliffe told the FT this month that Ineos was focused on “replacing the debt as it comes due” and would “not . . . be looking for new funding to do different things”, given conditions in the chemicals industry.
Inovyn makes chemicals including caustic soda and PVC, used widely in the construction sector, whose fortunes are linked closely to the wider economy. Demand for PVC has been muted in Europe, which accounted for 70 to 90 per cent of Inovyn’s PVC sales in 2024. The business is also highly exposed to electricity prices, which have soared in Europe since Russia’s invasion of Ukraine.
Its assets also include its Runcorn facility, which makes the chlorine that purifies 98 per cent of the UK’s drinking water. It is unclear whether this business will be included in the group’s disposal plans. Inovyn operates sites across Europe, with facilities in the UK, France, Italy, Germany, Belgium, Spain, Norway and Sweden.
Inovyn’s adjusted ebitda fell 41 per cent in 2024 — the most recent year for which figures are reported — to £348mn, against revenues of £3.1bn. Last year it announced the closure of two facilities in Germany, blaming “crippling energy and carbon costs, and a lack of tariff protection”.
S&P on Thursday downgraded Quattro’s debt deeper into junk territory, on estimates that its earnings had fallen in 2025, further raising leverage. The rating agency said that Inovyn’s “energy-intensive” PVC operations put “further pressure” on Quattro’s profit margins.
Ineos last year sold its composites assets to US private equity group KPS Capital Partners.
Ratcliffe’s stake in Ineos has made him one of the UK’s wealthiest individuals and in 2024 he became a co-owner of Manchester United. The billionaire, who lives in Monaco, apologised in February after he said the UK had been “colonised by immigrants”, drawing criticism from UK Prime Minister Sir Keir Starmer.
Ineos declined to comment.
Anthropic and OpenAI Take Different Stances On Disrupting Software Incumbents
Anthropic on Tuesday unveiled new details of how businesses could use its Claude Cowork AI software to access and use data stored in enterprise apps, including DocuSign, LegalZoom and Salesforce. Given the recent market jitters about the disruptive impact of AI on these software firms, you’d think the news would have depressed stocks even more.
Instead, stocks recovered a bit. Figma shares, for instance, were up 10%, while ServiceNow inched up 1.4% and Salesforce rose 4%. What’s going on? Possibly the actual details weren’t as bad as some investors had feared.
Indeed, Tuesday’s update seems to position Claude Cowork as a potential replacement for employees rather than the software they use. New AI tools will continue to use the software, which means businesses will still have to pay for it. (For more on what Cowork actually does, see this, this and this.)
It’s a delicate balance Anthropic seems keenly aware of. At an event announcing the new Cowork features, the AI firm looped in its head of economics, Peter McCrory, who said AI’s impact on labor would be “very uneven.” He said high-skilled workers would become more productive with AI while lower level data-entry workers could potentially be replaced.
Derek Hernandez, an analyst at PitchBook Data, said Anthropic’s Claude features for financial services tasks could put jobs in that field at risk.
“The disruption is seats at investment banks, seats in equity research, white collar jobs particularly on the lowest levels,” he said.
The tone Anthropic used is striking because it highlights partnerships, rather than competition, with software providers and risks to human employees. And inside Anthropic, employees continue to use all manner of traditional enterprise apps, as we’ve noted before. While that shouldn’t completely allay concerns from software providers—who of course benefit from larger workforces rather than shrinking ones—it contrasts somewhat with OpenAI’s public and private messaging about enterprise apps.
OpenAI’s Worker-Savings Calculation
To be sure, OpenAI has struck similar partnerships with enterprise firms like Salesforce, such as allowing ChatGPT customers to use Salesforce apps from the chatbot. But OpenAI told its investors in a meeting last week that its AI agents and future products would be capable of replacing software from tech firms including Salesforce, Workday, Adobe and Atlassian, my colleagues reported Friday.
As if to emphasize the point, OpenAI showed investors the revenues of those enterprise software firms, its own current revenue and the gargantuan revenue it projected to generate by 2030. (Of course, much of that revenue will come from OpenAI’s consumer ChatGPT business, not from enterprises per se.)
OpenAI leaders also told investors it believes the average worker using ChatGPT is saving about 50 minutes per day, translating to savings of about $50 a day per person. (For those calculations, OpenAI relied in part on estimates from Ark Invest, an OpenAI shareholder.) The business-friendly version of ChatGPT starts at $25 per worker per month, so OpenAI believes it’s only capturing a small fraction of the value it’s providing.
The private comments appear to confirm what many in the enterprise software world suspected when OpenAI announced its new “Frontier” AI product last month. It sees its technology as sitting on top of companies’ enterprise applications that store critical corporate data, exerting more influence on how businesses use and pay for software and AI.
Saas Strikes Back?
Meantime, many legacy software providers are staging a defense from the potential assault they face from OpenAI and its AI ilk. As we reported this morning, companies like ServiceNow and Microsoft are, predictably, aiming to convince customers that their software is more reliable than AI from labs like OpenAI and Anthropic. Other enterprise firms such as HubSpot are thinking about charging customers who want to use AI agents to tap their data stored in HubSpot’s systems.
“We are not a free data pipeline for everybody to take that information out,” HubSpot CEO Yamini Rangan told shareholders earlier this month.
The question now is how much power HubSpot and its ilk will have to slow down AI agents. The whole point of AI agents like OpenClaw is to take over a person’s computer and use apps the way that person would. Even Microsoft CEO Satya Nadella said in a 2024 podcast that he didn’t think his company would be able to block AI agents from using enterprise apps.
We’ll be closely following efforts from HubSpot and others to act as a kind of tollbooth operator, extracting what they can from customers using AI. We imagine this move will generate some fireworks.
Enterprise firms’ strongest defense against AI upstarts is their longstanding experience in complying with global data laws and privacy regulations—something that experimental AI products might not be able to handle for some time.
CrowdStrike Savings
It’s worth noting that most of the customers we’ve spoken to in recent weeks haven’t actually tried to replace enterprise software with AI. But some of them said they’re increasingly relying on agents to automate tasks using the raw data sitting in their existing systems, which is changing how employees interact with the software.
For instance, one cybersecurity executive told us he used an AI agent to avoid spending over $100,000 in annual fees for a CrowdStrike product that automated the process of managing employees’ accounts. The tool can flag suspicious logins, or find accounts that appear dormant—indicating an employee left the company but wasn’t properly offboarded—and automatically lock them down.
Instead of using the CrowdStrike product, which can cost up to hundreds of dollars per user per month, he found he could do the same thing by hooking up an AI agent from the startup Torq, powered by OpenAI and Anthropic models, to his company’s raw login data that was already being collected by its Microsoft software. The Torq agent automated the account “cleanup” at a much lower price. (Read our full story for more details.)
That doesn’t necessarily mean all older software products are at risk of losing money, especially if customers need to keep them around as data sources for AI agents. (CrowdStrike said in a statement that it lets such AI agents connect to its software so customers can use them together.) But in the nearer term, it could make software products feel less crucial to customers and their employees, who may instead come to see AI as the primary tool for work.
Why OpenAI, Anthropic Missed Their Own Gross Margin Forecasts
OpenAI has become increasingly optimistic about the revenue it expects to generate from both consumers and businesses, pushing up its projections for the next five years. Clouding those forecasts, however, are escalating cloud server costs that have outstripped revenue growth.
Those pressures were stark last year, when OpenAI’s gross margins fell to 33% from 40% in 2024, missing its own forecasts of 46%. Archrival Anthropic, which also recently raised its revenue forecasts, said in December it anticipated 2025 gross margins of 40%. While that’s a big improvement from a negative 94% gross margin in 2024, it was still 10 percentage points short of Anthropic’s earlier goal.
For both companies, one of the main culprits was a jump in inference costs, the payments they make to cloud providers to power AI models when customers talk to their chatbots or use models through an application programming interface.
OpenAI quadrupled its inference costs last year, to $8.4 billion, higher than the $6.6 billion it forecast last summer. It has told potential investors it had to buy more expensive access to servers due to higher than expected demand for its services.
Cloud providers typically charge a higher rate to rent their servers on demand, known as spot instances, rather than reserving them in advance.
Anthropic’s 2025 inference costs, meanwhile, were expected to rise more than three times to $2.7 billion, also higher than the company had projected, though it’s not clear why.
The fact that both companies’ margins worsened as more people used its services is notable, because the average cost of renting computing power fell throughout the year and the companies have consistently said they’re finding ways to run large AI models more efficiently.
One reason OpenAI’s margins likely slipped is that it spends a lot of money powering ChatGPT for nonpaying users. Of OpenAI’s 910 million weekly users, only about 5% are paying customers, I’ve reported. Last year, nearly half of its total inference costs, or $3.9 billion, supported those nonpaying users, versus $4.5 billion for paying users, the company’s financial forecasts show.
Another factor is the type of AI these companies run. Last year OpenAI introduced video generator Sora, which takes more server power to run than text-based queries, as well as reasoning models that require more computing power to calculate answers than traditional large language models do.
The company also allowed users to experiment with new compute-intensive features before eventually introducing user limits. Those features included the GPT-4o model, which became popular for generating images in the style of Studio Ghibli, a Japanese animation studio, according to a person with knowledge of the company’s efforts.
There’s a bright spot, however. OpenAI has gotten more efficient at serving paying users: Its compute margin—the revenue left after subtracting the cost of running AI models for those customers—was roughly 70% in October, an increase from about 52% at the end of last year and roughly 35% in January 2024.
OpenAI plans to generate more revenue from nonpaying users, mainly in the form of ads and e-commerce, and by selling subscriptions to more of those users. In January, for instance, it introduced an ad-supported ChatGPT subscription priced at roughly $5 to $8 per month worldwide.
This year, OpenAI expects to spend about 66% of its $14.1 billion in inference costs on serving paying customers. By the end of the decade, it estimates 94% of its roughly $85 billion in inference costs will support paying users. That year, 2030, it projects that its gross margin will rise to about 67%.
Still, the recent financial performance of OpenAI and Anthropic raises questions about how they will hit their respective targets of 60%-plus gross profit margins by the end of the decade, which would put them in the same ballpark as some of today’s best-of-class, publicly traded software companies.
OpenAI and Anthropic seem to have no problem getting investors to cover these costs. But eventually they’ll need to prove that the business these users generate more than offsets the costs of supporting them.
Anthropic Research Memo Shows Focus on Rogue Agents, Scheming Models
The Takeaway
- Of nearly 50 projects proposed for its research fellows, many involved agents
- Agents have contributed to Anthropic’s recent growth but come with risks of errant behavior
- The proposals give a rare look into Anthropic’s research focus
Puncturing the buzz over AI agents such as Anthropic’s Claude Code and the open-source project OpenClaw is the prospect that these agents could get tricked into revealing sensitive information such as a person’s banking information. In a sign of those concerns, Anthropic earlier this year singled out rogue agents as a topic of focus for its research fellows.
Anthropic’s staff proposed that the fellows train an agent to misbehave in certain circumstances—say, by writing code with security vulnerabilities. They also asked the researchers to create a benchmark for measuring how often agents fall prey to security issues, according to copies of the proposals seen by The Information. In total, Anthropic proposed that the fellows work on 49 projects, ranging from training Claude to win cybersecurity challenges to investigating Chinese open-source models, giving a rare look into the company’s research priorities.
The fellows work under a more senior researcher to advance Anthropic’s work on AI safety and security. That excludes some of itscritical research, such as developing new techniques to train more powerful frontier models. Although the fellows only pursued about half of the proposed projects, the proposals give a window into the topics Anthropic’s researchers have identified as important. That’s significant because at Anthropic and rivals such as OpenAI, Google DeepMind and xAI, research is the first step toward developing new products and applications, as well as the guardrails that give customers confidence in using them.
Last year, the fellows program was responsible for over half of the research that Anthropic’s alignment team, which works on catastrophic risks from AI, published in November and December, according to a spokesperson for the company. The fellows, who are often college or graduate students, spend four to six months pursuing research projects chosen by Anthropic’s employees and collaborators, such as staff at Berkeley, Calif.–based AI research organization Redwood Research.
The program is “a huge uplift to us research-wise and also helps us bring more people into the field,” said Ethan Perez, who leads much of Anthropic’s safety research and helped start its fellowship program.
Of the 49 projects that Anthropic staff and collaborators proposed for the program that started in January, 15 of them focused on security. These generally involve understanding security issues that arise with agents and coming up with ideas on how to patch them. Dozens of others set out to oversee and steer the actions of AI systems, including those that are potentially scheming against their users.
For instance, one project proposed using Claude Opus, Anthropic’s leading model, to reproduce attacks so the company can better defend against them. Currently, when Anthropic discovers a new exploit against its agents, its employees have to manually create an environment that reproduces the attack—for instance, a fake banking website that hijacks its agents. Instead, the researchers proposed using Claude Opus to generate its own version of these websites, which employees could use to train models so they don’t fall for the attacks.
Preventing hackers from misusing its agents is crucial for Anthropic’s business. It gained an early lead against rivals such as OpenAI with its coding agent, Claude Code, and its related agent for nontechnical work like responding to emails, Claude Cowork.
Revenue from Claude Code, which launched last February, recently reached an annualized pace of $2.5 billion, a figure that does not include Cowork, according to an Anthropic spokesperson. This growth helped attract investors, which poured $30 billion into the company, at a $350 billion valuation before the investment, earlier this month.
But regular reports of agents going awry—say, deleting a person’s inbox—could limit customers’ embrace of such agents, underscoring the need for safeguards. Already, Anthropic advises Cowork users to “monitor Claude for suspicious actions.” The difficulty of blunting such attacks has also presented obstacles for OpenAI.
Anthropic researchers also proposed several projects focused on Chinese AI models, such as one that involved replicating innovations from Chinese AI labs—although none of the recent fellows elected to work on those projects, said Perez. It’s not clear why they were more interested in other work.
Another nine projects focused on understanding the internal workings of AI models, a mainstay of Anthropic’s research and an area in which Anthropic is hiring rapidly. The projects include uncovering the math behind some of the AI models’ more bizarre behavior.
For instance, one project aims to understand “LLM mind viruses” such as the parasitic personas AI models have reportedly adopted, in which they become obsessed with spirals and convince humans to post strange messages on social media, spreading the “virus” to other AI models.
Pursuing such research has become so important to AI companies that they have offered hundreds of millions of dollars in compensation to top researchers. Even the Anthropic fellows are well paid, receiving $3,850 per week in the upcoming programs, which comes out to a salary of over $200,000 per year, according to an application for the program.
In addition to helping with core research areas, fellowship projects allow Anthropic to explore “more offbeat ideas” that could turn out to be important research directions, Perez said.
Key Takeaways From Trump’s State of the Union Address
The president defended his leadership in speech to Congress infused with patriotism
President Trump doubled down on tariffs in his State of the Union address, imposing 10% duties on imports for 150 days despite a Supreme Court ruling.
President Trump defended his economic policies, blaming Democrats for high costs, and attacked Democrats on immigration, calling for voter identification laws.
President Trump avoided direct attacks on the Supreme Court over its tariffs ruling and named Vice President JD Vance to lead a “war on fraud.”
WASHINGTON—President Trump delivered a State of the Union address to the nation, setting a record for the longest speech of its kind. Polls show him struggling to connect with the public on the economy and Republicans bracing for a challenging midterm election. In an address heavy on patriotism, Trump billed the past year as a “turnaround for the ages.”
Here’s a look at the main takeaways from Tuesday’s address before a joint session of Congress.
Doubled down on tariffs
Trump addressed the nation only days after the Supreme Court’s 6-3 ruling striking down his sweeping tariffs agenda and the president’s swift move to impose 10% tariffs on imports from all countries for 150 days. He said last Saturday that he would raise duties to 15% but hasn’t done so yet.
Trump said “everything was working well” and the U.S. was gaining revenue through tariffs. He bemoaned the Supreme Court’s ruling but said foreign countries and companies wanted to keep the tariffs in place.
The president said tariffs were “saving the country” through the “kind of money we’re taking in” and implausibly said tariff revenue could one day replace the nation’s income tax, “taking a great financial burden off the people that I love.” But Trump’s new tariff approach is premised on a legal authority never before used by a president for tariffs and is likely to be challenged in court.
Pulled punches on the Supreme Court
The stage was set for a confrontation between Trump and the Supreme Court after the president vented his frustration over the court’s tariffs ruling last week. He previously said the families of justices Neil Gorsuch and Amy Coney Barrett—whom he nominated during his first term—should be “embarrassed” they voted with the majority opinion written by Chief Justice John Roberts.
But there were no major blows. When Trump arrived and walked past the area where the justices were seated, he shook hands with the four justices in attendance, including Roberts and Coney Barrett. Gorsuch didn’t attend. During his speech, the president referred to the tariffs decision as an “unfortunate ruling,” but avoided attacking the justices by name.
A defense of the economy
Trump blamed Democrats for making the high cost of living an issue. “You caused that problem,” he said, eliciting a standing ovation from Republicans. “Their policies created high prices; our policies are rapidly ending them.”
Trump painted a rosy picture of declining costs on major goods—eggs, beef and fuel—though prices haven’t broadly fallen. He said he planned to address “the crushing cost of healthcare” by calling on Congress to codify the healthcare framework he released earlier in the year, which seeks to redirect federal subsidies from insurers to consumers. The plan hasn’t gained momentum in Congress.
The president also highlighted his steps to reduce the cost of prescription drugs. He touted efforts to stop private investment firms from buying residential homes and asked Congress to codify the measures.
Polls show voters are frustrated with the economy and stock markets have displayed volatility in recent weeks, though the Dow has steadily climbed since he took office. The president’s advisers have urged him to focus on efforts to make Americans’ lives more affordable as Democrats emphasize economic hardship on the campaign trail. “The roaring economy is roaring like never before,” the president said.
Iran: diplomacy or conflict?
Trump said the U.S. is in negotiations with Iran to halt their nuclear program and his preference is to reach a resolution through diplomacy. But he said, “we haven’t heard those secret words, ‘we will never have a nuclear weapon.’”
The president said the Iranian regime was the world’s largest sponsor of terrorism and had killed more than 30,000 protesters during recent unrest in the country.
The U.S. has moved a vast array of military equipment and aircraft into the Middle East in recent weeks and has been considering an attack against Iran. Trump said he has pursued a foreign policy underpinned by “peace through strength.”
Tough talk on immigration
Trump’s speech presented an unapologetic defense of his immigration policy. He attacked Democrats for allowing a “border invasion” and blamed them for cutting off funding to the Department of Homeland Security.
In one of the speech’s more heated moments, he asked everyone in the chamber to stand if they agreed with the proposition that the first duty of elected officials was to protect American citizens instead of immigrants that have entered the country illegally. The line drew silence from Democrats, most of whom remained seated, and elicited prolonged applause from Republicans who broke out in chants of “U.S.A.”
“You should be ashamed of yourself not standing up,” Trump told congressional Democrats. A camera showed Rep. Ilhan Omar (D., Minn.), a frequent Trump target who represents a Minneapolis area district, shouting at the president and responding at one point, “You should be ashamed.”
Trump called for laws requiring voter identification to vote, and claimed that Democrats opposed such measures because they “want to cheat” in elections.
War on fraud
Trump named Vice President JD Vance as the leader of the “war on fraud.” The vice president, who is seen as a likely candidate for president in 2028, has been a critic of Minnesota Gov. Tim Walz, over a welfare fraud scandal involving many residents of Somali descent, which the state and federal government is now investigating.
“He’ll get it done, and if we’re able to find enough of that fraud, we will actually have a balanced budget,” Trump claimed. Vance said last month he was assembling a major interagency task force to take on the issue.
Trump said he expects his administration will root out more fraud and mentioned several Democratic-led cities that would be scrutinized. Immigration and Customs Enforcement officers in December were deployed en masse to Minnesota on the heels the fraud scandal, touching off widespread protests.
Olympic glory
Early in his speech, Trump introduced the gold-medal winning U.S. men’s Olympic hockey team, who emerged from the gallery of the House chamber to chants of “U.S.A.! U.S.A.!” after their overtime victory over Canada Sunday.
Trump basked in the glow of America’s first men’s hockey Olympic gold since the 1980 “Miracle on Ice” and announced that the team’s goaltender, Connor Hellebuyck, would receive the Presidential Medal of Freedom, the nation’s highest civilian honor. Trump said the women’s hockey team, which also defeated Canada to win gold, would soon be visiting the White House.
How Should Your Suit Fit? A Guide For Men
Everything you need to know about five key elements, from pant breaks to sleeve lengths